murdoch research repository · coastal and shelf science (2013), doi: 10.1016/j.ecss.2013.03.014....

MURDOCH RESEARCH REPOSITORY

This is the author’s final version of the work, as accepted for publication following peer review but without the publisher’s layout or pagination.

The definitive version is available at http://dx.doi.org/10.1016/j.ecss.2013.03.014

French, B., Clarke, K.R., Platell, M.E. and Potter, I.C. (2013) An

innovative statistical approach to constructing a readily comprehensible food web for a demersal fish community.

Estuarine, Coastal and Shelf Science, 125 . pp. 43-56.

http://researchrepository.murdoch.edu.au/15135/

Copyright: © 2013 Elsevier Ltd.

It is posted here for your personal use. No further distribution is permitted.

http://dx.doi.org/10.1016/j.ecss.2013.03.014http://researchrepository.murdoch.edu.au/15135/

Accepted Manuscript

An innovative statistical approach to constructing a readily comprehensible food webfor a demersal fish community

Ben French, K. Robert Clarke, Margaret E. Platell, Ian C. Potter

PII: S0272-7714(13)00140-6

DOI: 10.1016/j.ecss.2013.03.014

Reference: YECSS 4101

To appear in: Estuarine, Coastal and Shelf Science

Received Date: 29 November 2012

Revised Date: 6 March 2013

Accepted Date: 16 March 2013

Please cite this article as: French, B., Clarke, K.R., Platell, M.E., Potter, I.C., An innovative statisticalapproach to constructing a readily comprehensible food web for a demersal fish community, Estuarine,Coastal and Shelf Science (2013), doi: 10.1016/j.ecss.2013.03.014.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service toour customers we are providing this early version of the manuscript. The manuscript will undergocopyediting, typesetting, and review of the resulting proof before it is published in its final form. Pleasenote that during the production process errors may be discovered which could affect the content, and alllegal disclaimers that apply to the journal pertain.

http://dx.doi.org/10.1016/j.ecss.2013.03.014

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

1

An innovative statistical approach to constructing a readily comprehensible food web for a

demersal fish community

Ben Frencha, K. Robert Clarkea,b, Margaret E. Platella,c and Ian C. Pottera* a Centre for Fish and Fisheries Research, Murdoch University, South Street, Murdoch, Western Australia, 6150 b Plymouth Marine Laboratory, Prospect Place, West Hoe, Plymouth PL1 3DH, United Kingdom c School of Environmental and Life Sciences, University of Newcastle, Brush Rd, Ourimbah, New South Wales, 2258 *Corresponding author [email protected] Tel: 61 (08) 9239 8801 Fax: 61 (08) 9239 8808

Abstract

Many food webs are so complex that it is difficult to distinguish the relationships between predators

and their prey. We have therefore developed an approach that produces a food web which clearly

demonstrates the strengths of the relationships between the predator guilds of demersal fish and

their prey guilds in a coastal ecosystem. Subjecting volumetric dietary data for 35 abundant

predators along the lower western Australia coast to cluster analysis and the SIMPROF routine

separated the various species x length class combinations into 14 discrete predator guilds.

Following nMDS ordination, the sequence of points for these predator guilds represented a ‘trophic’

hierarchy. This demonstrated that, with increasing body size, several species progressed upwards

through this hierarchy, reflecting a marked change in diet, whereas others remained within the same

guild. A novel use of cluster analysis and SIMPROF then identified each group of prey that was

ingested in a common pattern across the full suite of predator guilds. This produced 12 discrete

groups of taxa (prey guilds) that each typically comprised similar ecological/functional prey, which

were then also aligned in a hierarchy. The hierarchical arrangements of the predator and prey guilds

were plotted against each other to show the percentage contribution of each prey guild to the diet of

each predator guild. The resultant shade plot demonstrates quantitatively how food resources are

spread among the fish species and revealed that two prey guilds, one containing cephalopods and

teleosts and the other small benthic/epibenthic crustaceans and polychaetes, were consumed by all

predator guilds.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

2

Introduction

There has been an increasing and worldwide recognition of the need to adopt an ecosystem-

based approach to fisheries management (EBFM) in order that ecosystems, and thus the fisheries

they support, are sustained in a healthy state (Ecosystems Principles Advisory Panel, 1996; Bergen

Declaration, 2002; Essington and Punt, 2011; Espinoza-Tenorio et al., 2012). Such an approach

involves considering the ecosystem as a whole, rather than just the target species, and thus

represents a holistic approach that emphasises the importance of understanding the reciprocal

interactions of humans and marine resources (Pikitch et al., 2004; Curtin and Prellezo, 2010;

Dickey-Collas et al., 2010; Espinoza-Tenorio et al., 2012). In its report to the United States

Congress, the Ecosystem Principles Advisory Panel (1996) recommended that a Fisheries

Ecosystem Plan (FEP) should be developed and that this should involve a series of actions. One of

the eight suggested actions included the proposal that a conceptual model of the food web in an

ecosystem should be constructed, based on data for the predator and prey of each targeted species

over time. This would then permit the anticipated effects of the allowed harvest on predator-prey

dynamics to be addressed.

The production of a sound food web requires a thorough understanding of the trophic

interrelationships of the main fished and unfished species in that ecosystem. Such webs are

traditionally constructed using the trophic interactions between the various predators and their prey

and is typically based on analyses of gut contents and/or stable isotope ratios (Ecosystems

Principles Advisory Panel, 1996; de Ruiter et al., 2005; Field and Francis, 2006; Moloney et al.,

2011). When developed from gut content data, they are often represented by complex ‘spider-web’

or ‘birds-nest’ diagrams (e.g. Hori et al., 1993; Link, 2002). Consequently, they are often so

complex that they “conceal more than they reveal” and, as a result, fundamental patterns may be

obscured by the high level of detail (Raffaelli, 2000). The need to reduce the complexity of the

representation of the interactions between predators and their prey led many workers to combine

predator species into either functional groups (Raffaelli, 2000) or trophic guilds that comprise

species with similar prey (Root, 1967; Bulman et al., 2001; Reum and Essington, 2008) and thereby

reduce the number of entities within the food web. This thereby facilitates a clearer understanding

of the main aspects of the structure and function of ecosystems (Fulton et al., 2007) and the

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

3

potential for interspecific competition (Pianka, 1980). Scientists have also attempted to reduce the

complexity of food webs by decreasing the number of prey entities through, for example,

combining them into functional categories (e.g. Reum and Essington, 2008). The above efforts to

reduce complexity involve a degree of subjectivity regarding the level and extent to which the

predator and/or prey species are grouped, which has often varied among studies and thus hindered

comparisons between studies.

The dietary compositions of many fish species change as those species increase in body size

(Werner and Gilliam, 1984; Blaber and Bulman, 1987; Platell et al., 1998a, 2010; Shepherd and

Clarkson, 2001; Cocheret de la Morinière et al., 2003; French et al., 2012) and also sometimes

change with time of year (Jaworski and Ragnarsson, 2006; Lek et al., 2011; Schückel et al., 2011).

It is thus necessary to consider whether the details of the food web are influenced by the body sizes

of the various species and/or are related to season, recognising that although a number of species

may undergo size-related and/or seasonal changes, they may not all follow the same trends and

body size may thereby not exert an overall significant influence on the structure of the food web. In

a study of the guild structure of fishes in Puget Sound (USA), based on the diets of 21 species, the

individuals were separated into large and small fish, when data were available for both size groups,

and according to the season of sampling, i.e. autumn, summer and winter (Reum and Essington,

2008). That dietary study had the great advantage of identifying statistically the various groups of

predators that consume similar prey, through using the permutation-based SIMPROF test (Clarke

et al., 2008), which does not assume any a priori hypotheses as to which predators form a guild. In

the context of time of year, that study found no evidence that the structure of the overall food web

changed with season, which is consistent with the conclusions drawn from comparable detailed

studies of fish communities on the upper shelf of south-eastern Australia and the mid-slope of

southern Tasmania (Bulman et al., 2001; 2002).

The initial aim of this study was to produce a food web that illustrates the relationships

between the abundant demersal fish species and their prey on the lower west coast of Australia,

through employing the detailed quantitative dietary data that were derived from analyses of the gut

contents of those species in samples covering a wide size range of each species and each season

(Table 1). It soon became apparent that, as in numerous other studies, traditional approaches would

yield a complex food web that was not readily comprehensible and thus of immediate value to

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

4

managers and ecologists. We thus used an innovative multivariate approach, which involved the use

of SIMPROF, to identify statistically the various predator and prey guilds and thereby reduce, to a

manageable level, the number of groups required for constructing the food web. This approach,

which is still based on sound quantitative data and a series of objective statistical hypothesis tests,

enabled us to produce a food web in the form of a readily interpretable ‘shade plot’ that reveals the

magnitude of the trophic relationships between the fish predators and their prey.

1. Materials and methods

1.1. Sampling of fish and treatment of gut samples

The 35 demersal fish species, whose dietary data were used in the current study (Table 1),

were collected from coastal marine waters along the lower west coast of Australia between Lancelin

at ca 33°00 S and Cape Naturaliste at ca 33°30 S and in which these species are abundant. Each

species was sampled by one or more of the following methods: otter trawling, rod and line fishing,

long lining, gill netting, seine netting and spear fishing. The fish were placed on ice immediately

after capture and the whole fish, or the carcass and gut contents when the fish had been filleted,

were transported to the laboratory where they were frozen. The total length (TL) of each fish was

measured to the nearest 1 mm and, when the gut contained food, it was removed and placed in 70%

ethanol, except in the case of the larger guts which were first fixed in 10% formalin.

The dietary items in the guts of each fish were examined under a dissecting microscope and

identified to the highest taxonomic separation possible. A total of 468 different taxa were identified

in the gut contents of the 35 fish species. The percentage volumetric contribution of each dietary

taxon to the total volume of the stomach and/or intestinal contents (%V) was estimated visually

(Hynes, 1950; Hyslop, 1980). Unidentifiable material was not included in the analyses.

2.2. Structure of data

The dietary data, date of capture and total length of each individual of the 35 fish species

were entered into a common database. As most of the dietary items typically were not able to be

identified to species or genus, and frequently not to family, the dietary data for each individual were

aggregated to a higher taxonomic level, usually order. The total number of orders or other higher

taxa (47), subsequently referred to as prey taxa, was considered both manageable and appropriate

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

5

for retaining important information on the relationships between the dietary composition of each

species and its body size and time of year of capture.

The date of capture of each fish was assigned to the appropriate season, i.e. summer

(December to February), autumn (March to May), winter (June to August) or spring (September to

November). Length class intervals of 100 mm TL were chosen for all species, as they provided a

sufficient but not excessive number of guts for each length class interval of each species to facilitate

comparability in statistical analyses that involved intra- and inter-specific data for dietary

compositions. Total length classes in mm are as follows. 1 =

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

6

estimates of the species x length class group structuring; not to mention producing an unwieldy and

unreliable table of results. (4) Furthermore, the inclusion of season as a component of the trophic

guild structure, i.e. predator groups which have the same species at the same length in different

predator guilds, would increase markedly the complexity of the plots of the relationships between

the predators and their prey and thus reduce the effectiveness of the plots as a management tool for

deciding conservation methods etc. for key predators and their prey. The decision to exclude season

is consistent with the fact that, in detailed studies, the overall dietary composition of the fish

communities of Puget Sound (USA), the upper shelf of south-eastern Australia and the mid-slope of

southern Tasmania (Australia) did not change with season (Reum and Essington, 1988; Bulman

et al., 2001; 2002).

It is reiterated that every attempt was made to obtain dietary data for a length class of each

species from each season. If prey taxa are therefore important to a certain species x length class

group (predator guild) during a particular season, the seasonal effect will still constitute part of the

analysis determining that guild. Thus, the aim is to average the seasonal effects for good

management reasons, rather than ignoring them, and thus ultimately to produce a more robust and

parsimonious description of the food web.

2.3. Initial screening of dietary data

The data for all length class by season combinations for the 35 fish species, which contained

at least three replicate fish, were extracted from the common database. As the number of replicates

for each length class by season combination for each species varied greatly, the data set was

unbalanced. The dietary data were therefore subjected to the following iterative process to explore

whether this imbalance would influence the results. The volumetric contributions of the dietary

items to each length class by season combination for each species were square root transformed and

the resultant data employed to create a Bray-Curtis similarity matrix. A ‘distance among centroids’

matrix was calculated in PERMANOVA+ (Anderson et al., 2008), namely the distances between

the centres of gravity of selected groups of points within the full-dimensional ‘Bray Curtis space’,

in which points are located so that their inter-point distances (Euclidean) equate to Bray-Curtis

dissimilarities in the original space of the transformed data matrix. These selected groups

correspond to fish from each length class by season combination for each species.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

7

It can be argued that this ‘distances among centroids’ matrix is the optimal description of the

mean relationships among the dietary compositions of these groups. However, this matrix does have

the significant disadvantage that it loses the link to the original scale of measurement of the data

matrix, and is therefore not amenable to the subsequent, objective approach of defining higher-level

group structures within both the predator and prey taxa, using the SIMPROF routine (Clarke et al.,

2008) - see below. An alternative, which retains this especially important link, is to average the

(transformed) data matrix itself into these same groups of fish species by length class by season, but

this may have the potential to distort the true inter-group relationships because of the unbalanced

group sizes. This is a result of the well-known ‘species accumulation’ effect, in which averages

from larger numbers of replicates are likely to contain more species (here, prey taxa) and thus

artefactually generate additional dissimilarity between groups of different sizes. In order to examine

whether such distortion exists in this case, a simple model matrix was created using Euclidean

distances between the numbers of replicates in each group. From the RELATE routine in PRIMER

v6 (Clarke and Gorley, 2006), a Spearman correlation ρ was first calculated between this model

matrix and the Bray-Curtis dissimilarities computed from the averages of the square root

transformed dietary data for each group. A very weak relationship here (ρ < 0.2) is considered to

indicate that the lack of balance in the numbers of replicates making up the averages was potentially

not a confounding factor for subsequent analyses. As the first RELATE value exceeded 0.2, the

original data matrix was therefore re-examined to identify, for each species, any length class by

season combinations (groups) that contained only a small number (n) of replicates. Such

combinations were successively removed (n < 4, n < 5, etc) until the RELATE ρ value fell below

the designated threshold of 0.2.

In conjunction with the above threshold, the RELATE ρ statistic was then calculated

between the optimal ‘distances among centroids matrix’ and the Bray-Curtis dissimilarities based

on simple averaging of the transformed data, with a Spearman correlation approaching 0.9

considered to indicate a high degree of conformity between the information in these two matrices.

These combined criteria were satisfied by retaining, for every species, all length class by season

combinations that contained at least six replicates, the resulting RELATE correlations (ρ) between

centroid and average matrices then being 0.88, whilst the average and count matrices were

correlated at only the 0.19 level and the centroid and count matrices at only the 0.18 level.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

8

For the retained species by length class by season combinations, a Bray-Curtis matrix was

produced from the square-root transformed dietary volumetric data for all replicates in each

combination. This matrix was then subjected to a series of two-way crossed ANOSIM tests (Clarke,

1993), in which one factor (e.g. predator species) was crossed with the combined levels of the two

remaining factors (e.g. length class and season), thus removing the confounding effects of the latter.

This analysis was carried out separately for each of the three factors, removing the effects of the

other two, and the resultant global average R values were used to rank the factors in order of

importance in determining the assemblage of prey items in the diets. The factor found to be of least

importance, i.e. season, was ignored for subsequent analysis (see previous section for full rationale

for this exclusion) and thus the resulting calculations employed 112 combinations of species and

their length classes. This strengthened the number of replicates constituting each group, and the

results of re-analysis of the relationships between centroid and averaged matrices, i.e. ρ = 0.92, and

their relationship to sample size, i.e. ρ = 0.12 and ρ = 0.17, respectively, reinforced the validity of

working with the averaged matrix in the subsequent analyses.

2.4. Identification of predator guilds

The dietary compositions for the various species x 100 mm length class combinations for the

35 fish species were then grouped statistically into predator guilds, using an objective form of

cluster analysis. Specifically, the Bray-Curtis similarities from the above 112 group averages of

volumetric dietary data, now regarded as the ‘samples’ and considered to be effectively free from

sample-size bias, were subjected to hierarchical (Q-mode) cluster analysis using group-average

linking, and tested using the SIMPROF routine in PRIMER v6 (Clarke and Gorley, 2006; Clarke

et al., 2008). SIMPROF provides an objective means of defining, from the cluster dendrogram, the

sets of species x length-class combinations for which there is no evidence of the samples within

each set having any multivariate structure (e.g. further meaningful clustering of samples). This is

achieved by a hierarchical series of tests on the nodes of the dendrogram, progressing down the tree

to a finer level of classification of samples within a set only when there is evidence of remaining

multivariate structure. These SIMPROF sets therefore defined the ‘trophic guilds’ of predators, each

guild constituting different species and/or length-class combinations, such that similar diets are

found within each set, and are significantly different from those in other sets.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

9

A few of the resulting sets were outliers and, as they contained insufficient information for

credible inclusion in the ensuing guild analyses (e.g. they consisted of only one length class of one

species, and a low number of dietary samples), they were excluded from further consideration (see

Results). The relationships between the remaining 14 predator guilds were then examined in the

following two ways. Firstly, the Bray-Curtis resemblance matrix among samples (averaged data for

each predator species by length class combination) was input to a SIMPER analysis in PRIMER v6

(Clarke, 1993; Clarke and Gorley, 2006) giving, for each guild, the percentage contributions that

prey taxa made to the average within-guild similarity. From the full SIMPER tables, the prey taxa

principally typifying each predator guild were extracted.

Secondly, the same Bray-Curtis similarities were used to construct a ‘distances among

centroids’ matrix among the 14 predator guilds, using the PERMANOVA+ routine (Anderson et al.,

2008). A 2-dimensional non-metric MDS plot of the relationships among these 14 centroids was

then employed to display the gradient structure of trophic relationships among those various guilds.

Subsequently, summary measures, such as the number of predator species by length class

combinations making up each trophic guild, the total number of guts examined for these groups,

and the values for Simpson diversity of the average prey assemblage for each guild were displayed

as bubble plots on the 2-d nMDS ordination plot. The significance and extent to which the dietary

relationships amongst predator guilds are mirrored in Simpson evenness was quantified by the

RELATE routine (Clarke and Gorley, 2006), which, in this case, is a Spearman matrix correlation

between dietary Bray-Curtis dissimilarities and (Euclidean) distances between the values for

Simpson diversity, tested by permutation.

The main axis of the MDS ordination of predator guilds was also identified. Since axis

orientations are essentially arbitrary in MDS, this is defined as the first axis of a principal

component analysis of the 2-d MDS points, displayed in this case in the vertical direction, following

the usual convention for displaying hierarchies or gradients, with the guilds containing the largest

predators at the top of the plot.

2.5. Identification of prey guilds

The next step again involved SIMPROF, but this time to delineate each group of prey taxa (prey

guilds) within which the relative contributions to the diets of the trophic (predator) guilds were

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

10

similar. Therefore, after cluster analysis of the species x length class combinations and the

subsequent deletion of three outlying predator guilds, 44 of the original 47 prey taxa remain (see

Results, Fig. 1), as the three other prey taxa only occurred in the deleted predator guilds. A ‘species

resemblance’ matrix (Clarke and Warwick, 2001) can be defined between every pair of these prey

taxa by standardising the averaged data matrix (of 44 prey taxa by 14 predator guilds) over the

predator guilds, for each prey category (so that the values for each prey taxa sum to 100 over all

predator guilds), and then calculating Bray-Curtis similarities between prey taxa. (Note that this

method can be alternatively, and entirely equivalently described, as calculating Whittaker's Index of

Association (Whittaker, 1952) on the species of the original (unstandardised) matrix.) The resulting

resemblances reflect the viewpoint of the prey; i.e. what is the percentage breakdown of each prey

taxa across the predator guilds that consume it, and how similar are those percentage breakdowns

for the 44 different prey taxa? This species resemblance matrix was subjected to group-average

linked clustering (R-mode) in a manner similar to that used for the predator guilds (see earlier). In

conjunction with the cluster analysis, a further run of the SIMPROF routine (Clarke et al., 2008)

yields an objective grouping of the 44 prey taxa into ‘prey guilds’ (see Results for further details).

Prey taxa within each such guild are those for which the null hypothesis of indistinguishability in

their breakdown of percentage composition across the predator guilds cannot be rejected. Note that

such ‘species SIMPROF tests’ can be undertaken in PRIMER v6 but not straightforwardly, because

the default SIMPROF permutation procedure is not designed to carry out this novel analysis and

will permute the data matrix incorrectly. It thus requires temporary switching of the definition of

‘samples’ and ‘variables’ to obtain the correct permutation distributions (Somerfield and Clarke,

2011).

The resemblance matrix used for the cluster analysis of the prey taxa was then employed, as

described earlier for the predator guilds, to produce a nMDS plot of the ‘distances among centroids’

for the prey guilds and to determine the main axis of this plot. The common pattern of predation

within each prey guild is then illustrated by simple line plots showing the percentage consumption

of each prey taxa by each of the 14 predator guilds, with the predator and prey guilds each arranged

as in their order on the main axis of their respective nMDS ordination plots.

2.6. Food webs

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

11

A food web that linked the 44 prey taxa to the 112 predator species x length class

combinations would clearly be so complex that it would be uninformative. It is realistic, however,

to produce a web relating the ten prey guilds to the 14 predator guilds. For this purpose, the

volumetric percentage contributions of each prey taxa in a given prey guild are simply added, and

the resultant values averaged across all species x length class combinations in each of the predator

guilds. This enables a table to be constructed that provides the volumetric contribution of each prey

guild to the diet of a ‘typical’ member of each predator guild. These data were then square-root

transformed and rescaled so that, in an appropriate and clear visual manner, the lines linking the

various predator and prey guilds varied linearly in thickness on a food web plot in proportion to the

magnitude of the trophic interactions between those guilds.

Although the above food web comprises only cross-links between two discrete sets of

objects, i.e. predator guild and prey guild, and no internal links within those guilds, it is still very

complex. A more helpful and readily comprehensible representation of the relationships between

the predator and prey guilds is a ‘shade plot’, which uses the same square-root transformed

volumetric dietary data as employed for the above food web, but with rows and columns

representing the prey and predator guilds, respectively, and the depth of shading in each cell of this

two-way layout being linearly related on a continuous scale to the strength of the trophic interaction

in this second simpler food web.

The sequence of the predator and prey guilds in both the traditional food web and the food

web displayed as a shade plot follow those designated by their respective positions along the main

axis (vertical alignment) in their respective nMDS ordination plots (see earlier).

3. Results

3.1. Identifying predator guilds and their typifying prey species

The cluster dendrogram, derived from the Bray-Curtis similarity matrix constructed from the

volumetric dietary data for the length classes of each species, is shown in Fig. 1. Subjecting these

dietary data to SIMPROF separated the 112 species x length class combinations into 17 predator

guilds, designated as A to Q, which were significantly different from each other using a sequence of

P < 5% level tests, among which there were four outliers (Fig. 1). Although one of the outliers

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

12

(guild K) comprised a single species x length class combination, it contained as many as 37

replicates and was therefore considered a bona fide guild and thus retained for subsequent analyses.

The three other outliers (guilds A, C and J) each contained only one species x length class

combination and few replicates and were thus not included in subsequent analyses. There was thus

data for a total of 14 predator guilds for analysis.

On the ordination plot, derived from the volumetric dietary data for the above 14 predator

guilds, the points for those guilds followed a broadly downward progression from B at the top to I

at the bottom (Fig. 2). Major artefactual effects on this plot can be ruled out for the following

reasons. The number of species by length class combinations in each predator guild, as reflected in

the relative sizes of the bubbles for each guild in Fig. 3a, showed no overall tendency to change

consistently with its position on that ordination plot. Similarly, there was no evidence that the total

number of guts examined for dietary analyses varied with position on the same ordination plot

(Fig. 3b). Thus, in keeping with the earlier RELATE tests (see the Methods section 2.3. on Initial

Screening of Dietary Data), the order in which the predator guilds are distributed in the vertical axis

in Fig. 2 is related neither to the number of species by length class groups in each predator guild nor

to the number of individual guts in those guilds.

The vertical sequence of the 14 predator guilds in Fig. 2 is given in Table 2, commencing

with guild B and ending with predator guild I. This sequence progresses from the larger individuals

of the larger species, such as the teleosts Epinephelides armatus and Glaucosoma hebraicum and

the elasmobranchs Heterodontus portusjacksoni and Squatina australis (predator guilds B and D),

to the smallest individuals of four sillaginid species (predator guild O) and to smaller individuals of

Pseudocaranx georgianus and the small species Ammotretis elongatus (predator guild I).

The use of SIMPER demonstrated that the typifying prey taxa of the guilds at the top of

Table 2 (B, D and E) comprise the largest prey, i.e. teleosts and other decapods (mainly brachyuran

crabs), whereas those of predator guilds at the bottom of that table comprise the smaller prey, such

as cumaceans, amphipods and mysids. The data in Table 2 also emphasise that the predator guild of

the larger species can change markedly and progressively with increasing body size. This

phenomenon is exemplified by Pseudocaranx georgianus, with its predator guild shifting from I for

its smaller individuals, to F, near the top for its largest individuals (Table 2). Bubbles, whose sizes

represent the magnitude of the values for Simpson’s Diversity Index, were superimposed on the

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

13

points for the predator guilds in the ordination plot shown in Fig. 2 (Fig. 3c). The trends exhibited

by bubble size demonstrated that the diets were less diverse for predator guilds in the upper part of

the plot (B, D, E and F), which represented the larger individuals of the larger fish species, than for

all of those in the lower part of the plot and sometimes markedly so (guilds I, N, L and K) and

which represented the smaller individuals of large species and the smaller fish species. The apparent

pattern of increase in bubble size from top to bottom of the ordination plot is statistically established

by the RELATE test between the (Bray-Curtis) resemblance matrix for diets of the predator guilds

and the (Euclidean) distances between Simpson diversity values, which gives a matrix correlation of

ρ = 0.32, P < 1%.

3.2. Identifying prey guilds and their relationships to predator guilds

Cluster analysis of the volumetric contribution of each prey taxon to the diets of each

predator guild, expressed as a percentage of the total volumetric consumption of that prey taxon by

all predator guilds collectively, allied with the use of SIMPROF, yielded 12 groups (a-l) whose

compositions were significantly different from each other in a series of 5% level tests (Fig. 4).

Some prey guilds comprised relatively similar types of prey. For example, all groups of insects

were located in prey guild c, all cephalopods and teleosts in guild g, and guild l contained one

cluster comprising small epibenthic crustaceans, e.g. cumaceans, amphipods and mysids etc., and

another the two main groups of polychaetes, i.e. Errantia and Sedentaria (Fig. 4).

On the centroid ordination plot, derived from the same data as employed for the above

cluster analysis, the points for prey guilds e, f, d and g lie at the top, those for h, j, k and l in the

middle and those for i, c and b at the bottom, with prey guild a lying far to the left (Fig. 5). At one

extreme, prey guilds e, f and d comprised the largest of the sedentary prey that were consumed by

the 35 fish species, e.g. spatangoid echinoderms and archaeogastropod and mytiloid molluscs,

whereas, at the other extreme, prey guilds i, c and b comprised small planktonic crustaceans and

insect larvae.

The patterns displayed by the line plots in Fig. 6 emphasise that the relationships between

the percentage consumption of each prey taxon within each prey guild are similar. Thus, the prey

taxa in prey guilds e and f were consumed very largely only by one or both of predator guilds E and

F, whereas those in prey guilds c and b were ingested almost exclusively by one or both of predator

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

14

guilds G and K, which, in these cases, comprised the small individuals of sillaginid species and the

small species Atherinomorus ogilbyi and Spratelloides robustus (Table 2, Fig. 6). In contrast, prey

guild l was consumed by a wide range of predator guilds.

3.3. Food webs

Some trophic interactions can be clearly identified between certain predator and prey guilds

in the food web shown in Fig. 7, and particularly at the top and bottom of that web. Thus, for

example, the thickness of the lines relating predator guild B with the various prey guilds emphasise

that the members of this guild feed predominantly on prey guild g and likewise the members of

predator guild I feed largely on members of prey guild l. The trophic relationships are far more

difficult to detect, however, in the middle part of the food web, where there is extensive criss-

crossing of lines between many of the predator and prey guilds (Fig. 7).

The depth of the shading for the relationship between each predator guild and prey guild in

the shade plot shown in Fig. 8 reflects the magnitude of the interaction between those two guilds,

with the predator and prey guilds each being arranged in the sequences designated by the results of

the ordinations described earlier and shown in Figs 2 and 5, respectively. The trends emphasise that

the extent of the interaction between the prey guilds and the predator guilds broadly shifts in a

diagonal direction from top left to bottom right of the plot. Fig. 8 also illustrates very clearly that

some prey, such as those belonging to g and l, are consumed by the members of all predator guilds,

whereas others, such as those representing e, f and a, are ingested by only one or two predator

guilds. Furthermore, prey guilds h and k are fed on by predators in the centre of the hierarchy. The

plot also emphasises that predator guilds such as B and I fed on only three prey guilds, whereas, at

the other extreme, predator guild P fed on a wide spectrum of prey guilds (Fig. 8).

4. Discussion

4.1. Relationships between predator guilds and prey taxa

This study has used a range of statistical analyses and approaches to develop a food web that

can readily be used by scientists and managers to understand the strengths of the relationships

between a suite of abundant demersal fish predators and their prey in a coastal ecosystem. The

construction of this sound food web was facilitated by the availability of comprehensive

quantitative dietary data for a wide size range of 35 demersal fish species caught seasonally on the

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

15

lower west coast of Australia. The employment of the recently-developed SIMPROF technique

(Clarke et al., 2008) enabled the predator and prey guilds to be identified statistically and without

any a priori hypotheses, with the prey guilds being identified using an innovative version of this

SIMPROF test. The use of nMDS ordination enabled the hierarchical structure of both the predator

and prey guilds to be determined objectively and thus facilitate the matching of the components of

those two hierarchies in the form of a shade plot, which illustrates, in an effective and visual

manner, the magnitudes of the relationships between each predator guild and prey guild. It is

recognised that this shade plot focuses on those relationships and does not incorporate data for

lower levels in the food web, i.e. the relationships between primary consumers and primary

producers.

The statistical identification of those fish species x length class combinations, whose diets

were similar and differed from other such combinations, reduced the number of such combinations

in the data matrix (112) to a far more manageable number of predator guilds (14), while retaining

the resolution required for making meaningful dietary comparisons. The construction of these

predator guilds was thus not subjective and avoided the ad hoc methods, which, as pointed out by

Luczkovich et al. (2002), have frequently been used to aggregate predators into trophic guilds.

While the type of boot-strapping approach developed by Jaksic and Medel (1990), and used by

Garrison and Link (2000) in their dietary studies, also provides an objective method for

distinguishing between dietary groups, it produces only a single cut-off for the full data set, whereas

the use of cluster analysis with SIMPROF has the advantage of testing for significance between the

different species x length class combinations that represent the various nodes within the

dendrogram.

It was particularly notable that, when the centroids of the dietary data for the predator guilds

were subjected to nMDS ordination, the main axis of those guilds was aligned on the ordination plot

from the larger individuals of the largest fish species at one extreme and the smaller individuals of

the larger species and all of those of smallest species at the other. When that main axis was aligned

to the vertical, the composition of the prey changed progressively from those of the larger predators

at the top of the plot to those of the smaller predators at the bottom of the plot, thereby constituting

a trophic hierarchy. The larger individuals of the fish predators tended to feed predominantly on

other teleosts and other large prey, such as members of the Decapoda, and, in particular, brachyuran

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

16

crabs, while small fish ingested a wide range of small crustaceans, including amphipods, mysids,

cumaceans and carideans (see Platell et al., 1997, 1998a,b; 2010; Platell and Potter, 1998, 1999,

2001; Sommerville et al., 2011; French et al., 2012 for comprehensive dietary data for the separate

species). This trend was reflected in an increase in the diversity of the diet from the top to the

bottom of the hierarchy.

The hierarchical arrangement of the predator guilds, in combination with the distribution of

the length class groups for each predator species within those guilds, demonstrates that, as several

species of predator increase in body size, they progress sequentially upwards by at least one guild in

the trophic hierarchy and sometimes far more (Table 2). A particularly extreme example is provided

by the carangid Pseudocaranx georgianus, which belongs to predator guild I when small and thus

feeds mainly on cumaceans and amphipods, and to predator guild F when large and therefore feeds

predominantly on other decapods (mainly brachyurans) and teleosts. It was also noteworthy that the

two largest of the six sillaginids, Sillaginodes punctata and Sillago schomburgkii, underwent a

similar progressive upward shift in the trophic hierarchy from predator guild O when small to guild

P when of moderate size and finally to Q when large. Thus, the most important typifying prey taxa

were initially harpacticoid crustaceans, and then amphipod crustaceans and finally sedentary

polychaetes with the largest individuals (Table 2). These size-related shifts in the main prey taxa of

large species from one predator guild to one or more further guilds would reduce the potential for

intra-specific competition for food resources by these species. This conclusion parallels that drawn

by exploring the trends exhibited by the diets of individual species as they increase in size (Hyndes

et al., 1997; French et al., 2012), recognising that, in the case of Sillaginodes punctata, such

competition would also be reduced by the tendency for larger fish to move into deeper waters and

around reefs (Hyndes et al., 1998), a movement pattern exhibited by numerous fish species.

In contrast to the above trends, some larger species, such as Myliobatis australis and

Bodianus frenchii, remained throughout life in the same predator guild (F) and the same was very

largely true for Pagrus auratus, with the typifying prey species of this guild comprising other

decapods (mainly brachyurans) and teleosts. This lack of distinction between the predator guilds for

the various length classes of these large fish species is considered valid because the number of prey

taxa used was substantial (44). Indeed, that number, although similar to that of a recent compilation

of dietary data for 76 fishes in north-western Australia (Farmer and Wilson, 2011), was far greater

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

17

than the 8 and 26 employed in comparable studies by Reum and Essington (2008) and Akin and

Winemiller (2006), respectively. While it should be recognised that the overall compositions of the

diets of these species did change with increasing body size when using a finer taxonomic scale

(Platell et al., 2010; Sommerville et al., 2011; French et al., 2012), the use of those finer taxonomic

scales for the dietary categories in the present study would have produced a prohibitively large

number of predator guilds for the analyses employed in the current study and thus mitigated against

the construction of a readily comprehensible food web.

4.2. Food webs, including identification and characteristics of prey guilds

Until now, the discussion has largely focused on how food resources are partitioned among

demersal fish species on the lower west coast of Australia, taking into account the size of the fish.

The emphasis now shifts to exploring the ways in which food resources are shared among the

various fish predator guilds. This was achieved by identifying the various groups of prey taxa,

which had each been shown statistically to share common patterns of predation across one or more

predator guilds. This was achieved by using a novel ‘switching’ approach within SIMPROF

(R-mode analysis), which had the great advantage of reducing the number of 47 prey taxa in the

present study to a far more manageable number of prey guilds (12), thereby paralleling the benefits

of using SIMPROF to identify predator guilds (see above).

The prey taxa within each prey guild, which were objectively identified by the use of cluster

analysis with SIMPROF, showed a strong tendency to represent suites of prey with common

distinctive ecological/functional characteristics. For example, all cephalopod and teleost prey,

which are relatively large and mobile, are located in prey guild g, whereas prey guild b contained all

of the very small planktonic crustaceans, represented by the Notostraca, Calanoida and Cladocera.

Furthermore, the ‘largest’ of the prey guilds (l) comprised small benthic and epibenthic crustaceans

and the errant and sedentary polychaetes, which are not particularly mobile and live on or within the

substratum. Within prey guild f, the molluscs (mytiloids, mesogastropods, arcoids) and

echinoderms (clypeasteroids) are relatively large and immobile, and cirripedes and leptostracans are

amongst a multitude of taxa that live in or on structures created by mytiloids (e.g. Cinar et al., 2008;

Galkin and Goroslavskaya, 2008). Prey guild c contained all of the insects, represented by either

their larvae or adults. The larvae of the insects belonged in particular to the Tipulidae (Hourston

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

18

et al., 2004), whose pupae possess plastron-bearing spiracular gills and are found in saltwater

(Hinton, 1967), while the adults were represented by insects, such as those of the Formicidae, which

alight on the water surface (Hourston et al., 2004). These results emphasise that, in the food web for

the lower west coast of Australia, the members of each guild of demersal fish predators typically

feed on prey that occupy a particular ecological niche. There are, however, a few cases where the

basis for the distribution of taxa among guilds is not clear. For example, it is not evident why

prosobranchs and cubomedusae are present together in prey guild h, and why opisthobranchs and

phyrnophiurids occur together in prey guild j, in which they are the sole representatives. These

pairings are likely to reflect some commonality in terms of ecology or function, but which, due to a

paucity of data for these groups in south-western Australian waters, are not at present readily

apparent.

The conventional food web shown in Fig. 7 emphasises that such webs are still very

complex, even when, as in that figure, the data for the various predators and prey have been

aggregated into guilds. Thus, the relationships between these guilds could be clearly identified in

only a limited number of cases. In contrast, the relationships between predator and prey guilds, and

their relative magnitudes, as shown by variations in shading, can readily be discerned in the ‘shade

plot’ in Fig. 8, which matches the predator guilds against the prey guilds, in the hierarchical orders

determined from the nMDS ordinations shown in Figs 2 and 5, respectively. Thus, the large

predators at the apex of their trophic hierarchy can be seen to focus particularly on prey near the

apex of the prey hierarchy, which is towards the top left hand corner of the plot. In contrast, the

smaller individuals of large species and the smaller species towards the base of the predator

hierarchy concentrate on consuming prey towards the lower end of the prey hierarchy, which is

situated towards the lower right hand of the shade plot.

The trends exhibited by the locations and intensities of shading in Fig. 8 emphasise that

cephalopods, teleosts and other decapods (prey guild g) are consumed by all predator guilds.

However, they also demonstrate that these larger, more mobile and/or hard-bodied prey are most

important as a food source for large species, such as Aptychotrema vincentiana,

Glaucosoma hebraicum and Heterodontus portusjacksoni and particularly their larger individuals,

which belong to predator guilds B, D and E at the top of the predator hierarchy. It is also evident

that the small crustaceans and polychaetes, which live in or on the benthos, belong to the other prey

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

19

guild (l) that is consumed by all predator guilds. In contrast to the situation with prey guild g,

however, the members of prey guild l are a far more important food source to predator guilds at the

bottom of the trophic hierarchy and which include small species such as Lesueurina platycephala,

Atherinomorus ogilbyi, Pempheris klunzingeri, Lepidotrigla modesta and Ammotretis elongatus and

the small individuals of larger species such as Pseudocaranx georgianus (M, L, K, N and I).

Although the larger individuals of the suite of sillaginids (predator guilds Q and P) lie in the middle

of the predator guild hierarchy and feed on cephalopods, teleosts and decapods (prey guild g) and to

a greater extent small benthic crustaceans and polychaetes (prey guild l), they are distinguished

from other predator guilds by consuming a substantial collective volume of gastropods, small

bivalves and brittle stars (prey guilds h, j and k). Thus, while two prey guilds are consumed by all

predator guilds, the other prey guilds are typically ingested by at least three other predator guilds.

The food resources are consequently spread among and within the demersal fish species on the

lower west coast, thereby reducing the potential for inter- and intra-specific competition.

The production of a food web in the form of a shade plot, as shown in Fig. 8, will allow

managers and scientists to be able readily to visualise the trophic relationships between the main

commercial and recreational species and their prey and the magnitudes of those relationships. A

graphical representation of this form is particularly effective (compared with a table) in assimilating

the broad structure of predator-prey relationships and highlighting the major prey in the diets of the

various predator groups. This in turn will allow the key trophic links in the ecosystem to be

identified and thereby enable the effects of any perturbations in those relationships to be predicted.

Conversely, the influence of anthropogenic and other activities on a given fish species can be

predicted, when such activities are known clearly to have an effect on the abundances of the key

prey of that species. This would be especially important in the case of fish species that were

particularly selective in their choice of prey.

Acknowledgements

Our gratitude is expressed to Kris Parker for help with formatting the dietary database and to the

numerous research students and postdoctoral fellows at Murdoch University who contributed to the

dietary studies that provided the data used for analyses during the current investigation. Financial

support was provided by the Western Australian Marine Science Initiative, the Australian Fisheries

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

20

Research and Development Corporation and Murdoch University. Ben French was the recipient of a

Murdoch University PhD scholarship. Bob Clarke acknowledges his Adjunct Professorship at

Murdoch University and Honorary Research Fellowship at the Plymouth Marine Laboratory. The

authors are grateful to three anonymous referees for their thorough reviews and many suggestions

for improvements.

References

Akin, S., Winemiller, K.O., 2006. Seasonal variation in food web composition and structure in a

temperate tidal estuary. Estuaries and Coasts 29, 552-567.

Anderson, M.J., Gorley, R.N., Clarke, K.R., 2008. PERMANOVA+ for PRIMER: Guide to

software and statistical methods. PRIMER-E, Plymouth Marine Laboratory, Plymouth, UK,

214 pp.

Bergen Declaration, 2002. Fifth International Conference on the Protection of the North Sea. 20-21

March 2002. Bergen, Norway. 50 pp. Retrieved from

http://www.ospar.org/html_documents/ospar/html/bergen_declaration_final.pdf

Blaber, S.J.M., Bulman, C.M., 1987. Diets of fishes of the upper continental slope of eastern

Australia: context, calorific values, dietary overlap and trophic relationships. Marine

Biology 95, 345-356.

Boalch, G.T., 1987. Changes in the phytoplankton of the Western English Channel in recent years.

British Phycological Journal 22, 225–235.

Bulman, C., Althaus, F., He, X., Bax, N.J., Williams, A., 2001. Diets and trophic guilds of demersal

fishes of the south-eastern Australian shelf. Marine and Freshwater Research 52, 537-548.

Bulman, C.M., He, X., Koslow, J.A., 2002. Trophic ecology of the mid-slope demersal fish

community off southern Tasmania, Australia. Marine and Freshwater Research 53, 59–72.

Çinar, M.E., Katağan, T., Koçak, F., Öztürk, B., Ergen, Z., Kocatas, A., Önen, M., Kirkim, F.,

Bakir, K., Kurt, G., Dağli, E., Açik, S., Doğa, A., Özcan, T., 2008. Faunal assemblages of

the mussel Mytilus galloprovincialis in and around Alsancak Harbour (Izmir Bay, eastern

Mediterranean) with special emphasis on alien species. Journal of Marine Systems 71, 1-17.

Clarke, K.R., 1993. Non-parametric multivariate analyses of changes in community structure.

Australian Journal of Ecology 18, 117-143.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

21

Clarke K.R., Gorley, R.N., 2006. PRIMER v6: User manual/tutorial, PRIMER-E, Plymouth Marine

Laboratory, Plymouth, UK, 190 pp.

Clarke, K.R., Somerfield, P.J., Gorley, R.N., 2008. Testing of null hypotheses in exploratory

community analyses: similarity profiles and biota-environment linkage. Journal of

Experimental Marine Biology and Ecology 366, 56-69.

Clarke, K.R., Warwick, R.M., 2001. Change in marine communities. PRIMER-E, Plymouth,

Plymouth Marine Laboratory, 172 pp.

Cocheret de la Morinière, E., Pollux, B., Nagelkerken, I., Hemminga, M., Huiskes, A., van der

Velde, G., 2003. Ontogenetic dietary changes of coral reef fishes in the mangrove-seagrass-

reef continuum: stable isotopes and gut-content analysis. Marine Ecology Progress Series

246, 279–289.

Curtin, R., Prezello, R., 2010. Understanding marine ecosystem based management: a literature

review. Marine Policy 34, 823-830.

de Ruiter, P.C., Wolters, V., Moore, J.C., Winemiller, K.O., 2005. Food web ecology: playing

Jenga and beyond. Science 309, 68-71.

Dickey-Collas, M., Nash, R.D.M., Brunel, T., van Damme, C.J.G, Marshall, C.T., Payne, M.R.,

Corten, A., Geffen, A.J., Peck, M.A., Hatfield, E.M.C., Hintzen, N.T., Enberg, K., Kell,

L.T., Simmonds, E.J., 2010. Lessons learned from stock collapse and recovery of North Sea

herring: a review. ICES Journal of Marine Science 67, 1875–1886.

Ecosystems Principles Advisory Panel, 1996. Ecosystem-based fishery management. A report to

Congress by the Ecosystem Principles Advisory Panel (National Marine Fisheries Service).

54 pp. Retrieved from

http://www.marinenz.org.nz/documents/EPAP_2003_Ecobased_Fish_Man.pdf

Espinoza-Tenorio, A., Wolff, M., Taylor, M.H., Espejel, I., 2012. What model suits ecosystem-

based fisheries management? A plea for a structured modeling process. Reviews in Fish

Biology and Fisheries 22, 81–94.

Essington, T.E., Punt, A.E., 2011. Implementing ecosystem-based fisheries management: advances,

challenges and emerging tools. Fish and Fisheries 12, 123-124.

Farmer, B.M,, Wilson, S.F., 2011. Diet of finfish targeted by fishers in northwest Australia and the

implications for trophic cascades. Environmental Biology of Fishes, 91, 71-85.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

22

Field, J., Francis, R., 2006. Considering ecosystem-based fisheries management in the California

Current. Marine Policy 30, 552-569.

French, B., Platell, M.E., Clarke, K.R., Potter, I.C., 2012. Ranking of length-class, seasonal and

regional effects on dietary compositions of the co-occurring Pagrus auratus (Sparidae) and

Pseudocaranx georgianus (Carangidae). Estuarine, Coastal and Shelf Science 115, 309-325.

Fulton, E.A., Smith, A.D.M., Smith, D.C., 2007. Alternative management strategies for southeast

Australian Commonwealth fisheries: Stage 2: Quantitative management strategy evaluation.

CSIRO, Hobart, Tasmania. 405 pp.

Galkin, S.V., Goroslavskaya, E.I., 2008. Bottom fauna associated with mussel beds and alvinellid

communities in the hydrothermal field at 9° N of the East Pacific Rise. Oceanology 48,

509-516.

Garrison, L., Link, J., 2000. Dietary guild structure of the fish community in the northeast United

States continental shelf ecosystem. Marine Ecology Progress Series 202, 231–240.

Geoscience Australia, 2012. Sunrise and sunset results. Retrieved from

http://www.ga.gov.au/bin/geodesy/run/sunrisenset

Hanson, C.E., Pattiaratchi, C., Waite, A.M., 2005. Seasonal production regimes off south-western

Australia: influence of the Capes and Leeuwin Currents on phytoplankton dynamics. Marine

and Freshwater Research 56, 1011–1026.

Hinton, H.E., 1967. Plastron respiration in the marine fly Canace. Journal of the Marine Biological

Association UK 47, 319-327.

Hori, M., Gashagaza, M.N., Nshombo, M., 1993. Littoral fish communities in Lake Tanganyika:

irreplaceable diversity supported by intricate interactions among species. Conservation

Biology 7, 657–666.

Hourston, M., Platell, M.E., Valesini, F.J, Potter, I.C., 2004. Factors influencing the diets of four

morphological divergent fish species in nearshore marine waters. Journal of the Marine

Biological Association of the United Kingdom 84, 805-817.

Hyndes, G.A., Platell, M.E., Potter, I.C., 1997. Relationships between diet and body size, mouth

morphology, habitat and movements of six sillaginid species in coastal waters: implications

for resource partitioning. Marine Biology 128, 585–598.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

23

Hyndes, G.A., Platell, M.E., Potter, I.C., 1998. Age composition, growth, reproductive biology, and

recruitment of King George whiting, Sillaginodes punctata, in coastal waters of

southwestern Australia. Fishery Bulletin 96, 258–270.

Hynes, H.B.N., 1950. The food of freshwater sticklebacks (Gasterosteus aculeatus and Pygosteus

pungitius), with a review of methods used in studies of the food of fishes. Journal of Animal

Ecology 19, 36-58.

Hyslop, E.J., 1980. Stomach contents analysis – a review of methods and their application. Journal

of Fish Biology 17, 411-430.

Jaksić, F.M., Medel, R.G., 1990. Objective recognition of guilds: testing for statistically significant

species clusters. Oecologia 82, 87–92.

Jaworski, A., Ragnarsson, S., 2006. Feeding habits of demersal fish in Icelandic waters: a

multivariate approach. ICES Journal of Marine Science 63, 1682–1694.

Lek, E., Fairclough, D.V., Platell, M.E., Clarke, K.R., Tweedley, J.R., Potter, I.C., 2011. To what

extent are the dietary compositions of three abundant, co-occurring labrid species different

and related to latitude, habitat, body size and season? Journal of Fish Biology 78,

1913-1943.

Link, J., 2002. Does food web theory work for marine ecosystems? Marine Ecology Progress Series

230, 1-9.

Luczkovich, J.J., Ward, G.P., Johnson, J.C., Christian, R.R., Baird, D., Neckles, H., Rizzo, W.M.,

2002. Determining the trophic guilds of fishes and macroinvertebrates in a seagrass food

web. Estuaries 25, 1143-1163.

Moloney, C.L., St John, M.A., Denman, K.L., Karl, D.M., Koster, F.W., Sundby, S., Wilson, R.P.,

2011. Weaving marine food webs from end to end under global change. Journal of Marine

Systems 84, 106-116.

Pianka, E.R., 1980. Guild structure in desert lizards. Oikos 35, 194–201.

Pikitch, E.K., Santora, C., Babcock, E.A., Bakun, A., Bonfil, R., Conover, D.O., Dayton, P.,

Doukakis, P., Fluharty, D., Heneman, B., Houde, E.D., Link, J., Livingston, P.A., Mangel,

M., McAllister, M.K., Pope, J., Sainsbury, K.J., 2004. Ecosystem based fishery

management. Science 305, 346-347.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

24

Platell, M.E., Potter, I.C., 1998. Distributions, size compositions and diets of two abundant benthic

ambush-feeding teleosts in coastal waters of south-western Australia. Journal of the Marine

Biological Association of the United Kingdom 78, 587-608.

Platell, M.E., Potter, I.C., 1999. Partitioning of habitat and prey by abundant and similar-sized

species of the Triglidae and Pempherididae (Teleostei) in coastal waters. Estuarine, Coastal

and Shelf Science 48, 235-253.

Platell, M.E., Potter, I.C., 2001. Partitioning of food resources amongst 18 abundant benthic

carnivorous fish species in marine waters on the lower west coast of Australia. Journal of

Experimental Marine Biology and Ecology 261, 31-54.

Platell, M.E., Hesp, S.A., Cossington, S.M., Lek, E., Moore, S.E., Potter, I.C., 2010. Influence of

selected factors on the dietary compositions of three targeted and co-occurring temperate

species of reef fishes: implications for food partitioning. Journal of Fish Biology 76,

1255-1276.

Platell, M.E., Potter, I.C., Clarke, K.R., 1998b. Do the habitats, mouth morphology and diets of the

mullids Upeneichthys stotti and U. lineatus in coastal waters of south-western Australia

differ? Journal of Fish Biology 52, 398-418.

Platell, M.E., Potter, I.C., Clarke, K.R., 1998a. Resource partitioning by four species of

elasmobranchs (Batoidea: Urolophidae) in coastal waters of temperate Australia. Marine

Biology 131, 719-734.

Platell, M.E., Sarre, G.A., Potter, I.C., 1997. The diets of two co-occurring marine teleosts,

Parequula melbournensis and Pseudocaranx wrighti, and their relationships to body size

and mouth morphology, and the season and location of capture. Environmental Biology of

Fishes 49, 361-376.

Raffaelli, D., 2000. Trends in research on shallow water food webs. Journal of Experimental Marine

Biology and Ecology 250, 223-232.

Reum, J.C.P., Essington, T.E., 2008. Seasonal variation in guild structure of the Puget Sound

demersal fish community. Estuaries and Coasts 31, 790-801.

Root, R.B., 1967. The niche exploitation pattern of the blue-grey gnatcatcher. Ecological

Monographs 37, 317-350.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

25

Schafer, L.N., Platell, M.E., Valesini, F.J, Potter, I.C., 2002. Comparisons between the influence of

habitat type, season and body size on the dietary compositions of fish species in nearshore

marine waters. Journal of Experimental Marine Biology and Ecology 278, 67-92.

Schückel, S., Sell, A., Kröncke, I., Reiss, H., 2011. Diet composition and resource partitioning in

two small flatfish species in the German Bight. Journal of Sea Research 66, 195-204.

Shepherd, S.A., Clarkson, P.S., 2001. Diet, feeding behaviour, activity and predation of the

temperate blue-throated wrasse, Notolabrus tetricus. Marine and Freshwater Research

52, 311-322.

Somerfield, P.J., Clarke, K.R., 2011. A statistical method to identify coherent groups of species

responding to spatio-temporal environmental gradients. Abstracts of the 9th International

Temperate Reefs Symposium, University of Plymouth, UK, 175 pp.

Sommerville, E., Platell, M.E., White, W.T., Jones, A.A., Potter, I.C., 2011. Partitioning of food

resources by four abundant, co-occurring elasmobranch species and the relationships

between diet and season, body size, region and feeding mode. Marine and Freshwater

Research 62, 54-65.

Werner, E.E., Gilliam, J.F., 1984. The ontogenetic niche and species interaction in size-related

populations. Annual Review of Ecology and Systematics 15, 393–425.

Whittaker, R.H., 1952. A study of summer foliage insect communities in the Great Smoky

Mountains. Ecological Monographs 22, 1-44.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

26

List of Figures

Figure 1 ‘Q-mode’ cluster dendrogram, derived from the Bray-Curtis similarity matrix constructed

from the volumetric dietary data for the length classes of each fish species for which there were

such data. The thick lines designate the species x length class combinations that were separated by

SIMPROF into a series of groups (predator guilds) whose dietary compositions differed. Note that

three of the four outliers (A, C and J) contained a single species x length class combination and

were not included in further analysis (see text for full rationale). Full generic names for each fish

species are provided in Table 2.

Figure 2 Centroid nMDS ordination plots of predator guilds, derived from a Bray-Curtis matrix of

the volumetric contributions of the prey taxa to each ‘sample’ (species x length class combination)

within the various predator guilds. Length classes are grouped in 100 mm TL intervals from

1 =

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

27

Figure 6 Line plots showing the pattern of consumption of each prey taxon, relative to its total

consumption, across the 14 predator guilds. The predator and prey guilds are both listed according

to their order on the vertical axes of their respective nMDS ordination plots (see Figs 2 and 5).

Figure 7 Traditional food web showing the trophic linkages between the predator and prey guilds.

The thickness of the links represent the relative strengths of the relationships.

Figure 8 A shade plot showing the relative magnitudes of the trophic linkages between the predator

and prey guilds, with the total consumption of all members of prey guild ‘x’ making up percentage

p of the diet of the average member of predator guild ‘X’, where the strength of the grey shading

represents the value of p (see shade legend, lower left), ranging from p=0 (white) to p=100%

(black). Note that those cells with only slight shading are delineated by faint grey lines.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

28

Table 1 The 35 demersal fish species whose diets were used to explore the trophic relationships between fish species and their prey on the lower west coast of Australia, together with the relevant publications or data sources.

Families Species Publications

Elasmobranchs Heterodontidae Heterodontus portusjacksoni Sommerville et al. (2011) Myliobatidae Myliobatis australis Sommerville et al. (2011) Rhinobatidae Aptychotrema vincentiana Sommerville et al. (2011) Squatinidae Squatina australis Sommerville et al. (2011) Urolophidae Trygonoptera mucosa Platell et al. (1998a)

Trygonoptera personata Platell et al. (1998a) Urolophus lobatus Platell et al. (1998a) Urolophus paucimaculatus Platell et al. (1998a) Teleosts

Atherinidae Atherinomorus ogilbyi Hourston et al. (2004) Carangidae Pseudocaranx georgianus French et al. (2012) Pseudocaranx wrighti Platell et al. (1997) Clupeidae Spratelloides robustus Schafer et al. (2002) Gerreidae Parequula melbournensis Platell et al. (1997) Glaucosomatidae Glaucosoma hebraicum Platell et al. (2010) Labridae Bodianus frenchii Platell et al. (2010) Leptoscopidae Lesueurina platycephala Hourston et al. (2004) Mullidae Upeneichthys lineatus Platell et al. (1998b) Upeneichthys stotti Platell et al. (1998b) Pempherididae Parapriacanthus elongatus Platell and Potter (1999) Pempheris klunzingeri Platell and Potter (1999) Platycephalidae Platycephalus longispinis Platell and Potter (1998) Pleuronectidae Ammotretis elongatus Hourston et al. (2004) Pseudorhombus jenynsii Schafer et al. (2002) Scorpaenidae Maxillicosta scabriceps Platell and Potter (1998) Serranidae Epinephelides armatus Platell et al. (2010) Sillaginidae Sillaginodes punctata Hyndes et al. (1997)

and Platell (unpublished data) Sillago burrus Hyndes et al. (1997) Sillago robusta Hyndes et al. (1997) Sillago schomburgkii Hourston et al. (2004) Sillago vittata Hyndes et al. (1997) Sillago bassensis Hyndes et al. (1997) Sparidae Pagrus auratus French et al. (2012) Rhabdosargus sarba Ang (unpublished data) Triglidae Lepidotrigla modesta Platell and Potter (1999)

Lepidotrigla papilio Platell and Potter (1999)

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

29

Table 2 The predator guilds identified among the 35 demersal fish species by SIMPROF, together with their typifying prey taxa and the percentage contributions made by each of those categories to the average similarity of the dietary composition of each predator guild (as identified by SIMPER). Note that each predator guild comprises groups of species x length class combinations. Length classes in mm are as follows. 1 = < 100, 2 = 100-199, 3 = 200-299, 4 = 300-399, 5 = 400-499, 6 = 500-599, 7 = 600-699, 8 = 700-799, 9 = 800-899 and 10 = 900-999.

Predator species

Length class (mm)

Predator guild

Prey taxa Percentage similarity

contribution

Aptychotrema vincentiana 10 Teleostei 97 Epinephelides armatus 3-5 Glaucosoma hebraicum 4,7-9 Heterodontus portusjacksoni 3,9 Squatina australis 3-10

B

Aptychotrema vincentiana 8,9 Teleostei 62 Epinephelides armatus 2 Other Decapoda 37 Heterodontus portusjacksoni 4,10

D

Pagrus auratus 6 Teleostei 67 Glaucosoma hebraicum 5,6

E Other Decapoda 22

Aptychotrema vincentiana 3-7 Other Decapoda 67 Bodianus frenchii 2-5 Teleostei 18 Glaucosoma hebraicum 1,2 Heterodontus portusjacksoni 7 Maxillicosta scabriceps 1,2 Myliobatis australis 2-5,7 Pagrus auratus 1-3,7-9 Platycephalus longispinis 2,3 Pseudocaranx georgianus 3-5 Pseudorhombus jenynsii 1,2 Rhabdosargus sarba 2 Upeneichthys lineatus 3

F

Sillaginodes punctata 4 Sedentaria 40 Sillago bassensis 3 Errantia 31 Sillago burrus 2,3 Sillago schomburgkii 4 Sillago vittata 2,3

Trygonoptera mucosa 2-4

Q

Pseudocaranx wrighti 1,2 Other Decapoda

37 Rhabdosargus sarba 3 Amphipoda 23

Upeneichthys lineatus 2

H

Tanaidacea 20

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

30

Table 2 continued.

Predator species Length

class (mm) Predator

guild Prey taxa

Percentage similarity

contribution

Parequula melbournensis 1,2 Amphipoda 40 Sillaginodes punctata 2,3 Errantia 31 Sillago bassensis 2 Sillago robusta 2 Sillago schomburgkii

2,3

P

Atherinomorus ogilbyi 1 Calanoida 69 Sillago bassensis 1 Amphipoda 14 Spratelloides robustus 1

G

Cladocera 11

Sillaginodes punctata 1 Harpacticoida 48 Sillago burrus 1 Errantia 19 Sillago schomburgkii 1 Amphipoda 16 Sillago vittata 1

O

Lesueurina platycephala 1 Amphipoda 39 Upeneichthys stotti 2 Mysidacea 21 Cumacea 14

M

Isopoda 12

Trygonoptera personata 2,3 Errantia 25 Urolophus paucimaculatus 3 Amphipoda 20 Caridea 19 Sedentaria 14

L

Mysidacea 13

Atherinomorus ogilbyi 2 K Amphipoda 85

Lepidotrigla modesta 1,2 Mysidacea 31 Lepidotrigla papilio 1,2 Amphipoda 27 Parapriacanthus elongatus 1,2 Cumacea 14 Pempheris klunzingeri 1,2 Caridea 11

Urolophus lobatus 2,3 Urolophus paucimaculatus 2

N

Ammotretis elongatus 1 Cumacea 56 Pseudocaranx georgianus 2

I Amphipoda 44

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

Glaucosoma hebraicum 4

Epinephelides armatus 4

Squatina australis 3







Aptychotrema vincentiana 10


Heterodontus portusjacksoni 3







Pseudocaranx georgianus 6







Pagrus auratus 6


Pagrus auratus 7

Pagrus auratus 2

Pagrus auratus 3



Myliobatis australis 2

Maxillicosta scabriceps 1

Maxillicosta scabriceps 2


Platycephalus longispinis 2

Pseudorhombus jenynsii 1


Platycephalus longispinis 3

Pseudorhombus jenynsii 2

Pagrus auratus 1



Bodianus frenchii 2

Bodianus frenchii 5

Bodianus frenchii 3

Bodianus frenchii 4

Rhabdosargus sarba 2






Pagrus auratus 8







Pagrus auratus 9

Spratelloides robustus 1

Atherinomorus ogilbyi 1

Sillago bassensis 1

Pseudocaranx wrighti 1

Pseudocaranx wrighti 2

Rhabdosargus sarba 3


Ammotretis elongatus 1


Upeneichthys stotti 1

Atherinomorus ogilbyi 2

Urolophus paucimaculatus 3

Trygonoptera personata 2

Trygonoptera personata 3

Lesueurina platycephala 1

Upeneichthys stotti 2

Pempheris klunzingeri 2

Urolophus paucimaculatus 2

Parapriacanthus elongatus 1

Parapriacanthus elongatus 2

Pempheris klunzingeri 1

Urolophus lobatus 2

Urolophus lobatus 3

Lepidotrigla modesta 1

Lepidotrigla modesta 2

Lepidotrigla papilio 1

Lepidotrigla papilio 2

Sillago vittata 1

Sillaginodes punctata 1

Sillago burrus 1

Sillago schomburgkii 1

Sillago robusta 2

Sillago bassensis 2





Parequula melbournensis 1

Parequula melbournensis 2



Trygonoptera mucosa 2



Sillago vittata 2

Sillago burrus 3

Sillago burrus 2

Sillago bassensis 3

Sillago vittata 3

100 80 60 40 20 0

Similarity (%)

A

B

C

D

E

F

M

N

L

P

G

IJ

H

K

O

Q

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

2D Stress: 0.09

DE

B

F

G

H

I

K LN

M O

P

Q

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

(b)

(a)

(c)

DE

B

F

G

H

I

K LN

M O

P

Q

2

8

14

20

2D Stress: 0.09

DE

F

I

KN

M

P

Q

200

800

1400

2000

2D Stress: 0.09

B

F

G

H

I

K LN

M O

P

Q

0.02

0.24

0.42

0.80

2D Stress: 0.09

DE

B

H

G

LO

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

Other Insecta

100 80 60 40 20 0

Similarity (%)

Orectolobiformes

Notostraca

Calanoida

Cladocera

Hymenoptera

Conchostraca

Coleoptera

Hemiptera

Diptera

Archaeogastropoda

Spatangoida

Mytiloida

Clypeasteroida

Other Leptostraca

Mesogastropoda

Arcoida

Cirripedia

Teuthida

Octopoda

Sepioidea

Other Decapoda

Teleostei

Prosobranchia

Cyclopoida

Cubomedusae

Dentaliida

Harpacticoida

Nebaliidae

Opisthobranchia

Phrynrophiurida

Solemyoida

Veneroida

Anaspidea

Cephalaspidea

Cumacea

Amphipoda

Tanaidacea

Caridea

Isopoda

Mysidacea

Stomatopoda

Errantia

Sedentaria

a

b

c

de

f

g

h

i

j

k

l

Very small

zooplanktonic

crustaceans

All insects

Small epibenthic

and benthic

crustaceans and

polychaetes

All cephalopods

and teleosts

Large molluscs

and echinoderms

Very small

epibenthic crustaceans

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

2D Stress: 0.14

l k

d

f

bc

h

g

i

j

a

e

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

i

l

k

j

a

d

f

e

10%

100%

Spatangoida

Arcoida

Cirripedia

Clypeasteroida

Mesogastropoda

Mytiloida

Other Leptostraca

Archaeogastropoda

Octopoda

Other Decapoda

Sepioidea

Teleostei

Teuthida

Orectolobiformes

Cubomedusae

Cyclopoida

Dentaliida

Prosobranchia

Opisthobranchia

Phyrnophiurida

Anaspidea

Cephalaspidea

Solemyida

Veneroida

Amphipoda

Caridea

Cumacea

Errantia

Isopoda

Mysidacea

Sedentaria

Stomatopoda

Tanaidacea

Harpacticoida

Nebaliidae

Coleoptera

Conchostraca

Diptera

Hemiptera

Hymenoptera

Other Insecta

Calanoida

Cladocera

Notostraca

Prey guild

h

g

K

N

I

L

M

O

G

P

H

F

D

B

Q

Aptychotrema vincentiana

Epinephelides armatus

Glaucosoma hebraicum

Heterodontus portusjacksoni

Squatina australis


Epinephelides armatus


Pagrus auratus



Bodianus frenchii



Maxillicosta scabriceps

Myliobatis australis

Pagrus auratus

Platycephalus longispinis

Pseudocaranx georgianus

Pseudorhombus jenynsii

Rhabdosargus sarba

Upeneichthys lineatus

Sillaginodes punctata

Sillago bassensis

Sillago burrus

Sillago schomburgkii

Sillago vittata

Trygonoptera mucosa

Pseudocaranx wrighti

Rhabdosargus sarba

Upeneichthys lineatus

Parequula melbournensis


Sillago bassensis

Sillago robusta


Atherinomorus ogilbyi

Sillago bassensis

Spratelloides robustus


Sillago burrus


Sillago vittata

Lesueurina platycephala

Upeneichthys stotti

Trygonoptera personata

Urolophus paucimaculatus

Atherinomorus ogilbyi

Lepidotrigla modesta

Lepidotrigla papilio

Parapriacanthus elongatus

Pempheris klunzingeri

Urolophus lobatus

Urolophus paucimaculatus

Ammotretis elongatus

Pseudocaranx georgianus

10

3-5

4,7-9

3,9

3-10

8,9

2

4,10

6

5,6

3-7

2-5

1,2

7

1,2

2-5,7

1-3,7-9

2,3

3-5

1,2

2

3

4

3

2,3

4

2,3

2-4

1,2

3

2

1,2

2,3

2

2

2,3

1

1

1

1

1

1

1

1

2

2,3

3

2

1,2

1,2

1,2

1,2

2,3

2

1

2

Predator guild

E

c

b

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPTe f d g a h j k l i c b

Spata

ngoid

a

Arc

oid

a

Cirripedia

Cly

peaste

roid

a

Mesogastr

opoda

Mytilo

ida

Oth

er

Lepto

str

aca

Arc

haeogastr

opoda

Octo

poda

Oth

er

Decapoda

Sepio

idea

Tele

oste

i

Teuth

ida

Ore

cto

lobiform

es

Cubom

edusae

Cyclo

poid

a

Denta

liida

Pro

sobra

nchia

Opis

thobra

nchia

Phyrn

ophiu

rida

An

asp

ide

a

Ce

ph

ala

sp

ide

a

So

lem

yid

a

Ve

ne

roid

a

Am

ph

ipo

da

Ca

rid

ea

Cu

ma

ce

a

Err

an

tia

Iso

po

da

Mysid

ace

a

Se

de

nta

ria

Sto

ma

top

od

a

Ta

na

ida

ce

a

Ha

rpa

ctico

ida

Ne

ba

liid

ae

Co

leo

pte

ra

Co

nch

ostr

aca

Dip

tera

He

mip

tera

Hym

en

op

tera

Oth

er

Inse

cta

Ca

lan

oid

a

Cla

do

ce

ra

No

tostr

aca

jadfe hg

ilk c b

Ap

tych

otr

em

a v

ince

ntia

na

Ep

ine

ph

elid

es a

rma

tus

Gla

uco

so

ma

he

bra

icu

m

He

tero

do

ntu

s p

ort

usja

ckso

ni

Sq

ua

tin

a a

ustr

alis

Ap

tych

otr

em

a v

ince

ntia

na

Ep

ine

ph

elid

es a

rma

tus

He

tero

do

ntu

s p

ort

usja

ckso

ni

Pa

gru

s a

murdoch research repository · coastal and shelf science (2013), doi: 10.1016/j.ecss.2013.03.014....

Documents